IEEE TRANSACTIONS ON POWER SYSTEMS, VOL. 19, NO. 2, MAY 2004
1387
Definition and Classification
of Power System Stability
IEEE/CIGRE Joint Task Force on Stability Terms and Definitions
Prabha Kundur (Canada, Convener), John Paserba (USA, Secretary), Venkat Ajjarapu (USA), Göran Andersson
(Switzerland), Anjan Bose (USA) , Claudio Canizares (Canada), Nikos Hatziargyriou (Greece), David Hill
(Australia), Alex Stankovic (USA), Carson Taylor (USA), Thierry Van Cutsem (Belgium), and Vijay Vittal (USA)
Abstract—The problem of defining and classifying power
system stability has been addressed by several previous CIGRE
and IEEE Task Force reports. These earlier efforts, however,
do not completely reflect current industry needs, experiences
and understanding. In particular, the definitions are not precise
and the classifications do not encompass all practical instability
scenarios.
This report developed by a Task Force, set up jointly by the
CIGRE Study Committee 38 and the IEEE Power System Dynamic
Performance Committee, addresses the issue of stability definition
and classification in power systems from a fundamental viewpoint
and closely examines the practical ramifications. The report aims
to define power system stability more precisely, provide a systematic basis for its classification, and discuss linkages to related issues
such as power system reliability and security.
Index Terms—Frequency stability, Lyapunov stability, oscillatory stability, power system stability, small-signal stability, terms
and definitions, transient stability, voltage stability.
I. INTRODUCTION
P
OWER system stability has been recognized as an important
problem for secure system operation since the 1920s [1], [2].
Many major blackouts caused by power system instability have
illustrated the importance of this phenomenon [3]. Historically,
transient instability has been the dominant stability problem on
most systems, and has been the focus of much of the industry’s
attention concerning system stability. As power systems have
evolved through continuing growth in interconnections, use of
new technologies and controls, and the increased operation in
highly stressed conditions, different forms of system instability
have emerged. For example, voltage stability, frequency stability
and interarea oscillations have become greater concerns than
in the past. This has created a need to review the definition and
classification of power system stability. A clear understanding
of different types of instability and how they are interrelated
is essential for the satisfactory design and operation of power
systems. As well, consistent use of terminology is required
for developing system design and operating criteria, standard
analytical tools, and study procedures.
The problem of defining and classifying power system stability is an old one, and there have been several previous reports
Manuscript received July 8, 2003.
Digital Object Identifier 10.1109/TPWRS.2004.825981
on the subject by CIGRE and IEEE Task Forces [4]–[7]. These,
however, do not completely reflect current industry needs, experiences, and understanding. In particular, definitions are not
precise and the classifications do not encompass all practical instability scenarios.
This report is the result of long deliberations of the Task Force
set up jointly by the CIGRE Study Committee 38 and the IEEE
Power System Dynamic Performance Committee. Our objectives are to:
• Define power system stability more precisely, inclusive of
all forms.
• Provide a systematic basis for classifying power system
stability, identifying and defining different categories, and
providing a broad picture of the phenomena.
• Discuss linkages to related issues such as power system
reliability and security.
Power system stability is similar to the stability of any
dynamic system, and has fundamental mathematical underpinnings. Precise definitions of stability can be found in the
literature dealing with the rigorous mathematical theory of
stability of dynamic systems. Our intent here is to provide a
physically motivated definition of power system stability which
in broad terms conforms to precise mathematical definitions.
The report is organized as follows. In Section II the definition of Power System Stability is provided. A detailed
discussion and elaboration of the definition are presented.
The conformance of this definition with the system theoretic
definitions is established. Section III provides a detailed classification of power system stability. In Section IV of the report the
relationship between the concepts of power system reliability,
security, and stability is discussed. A description of how these
terms have been defined and used in practice is also provided.
Finally, in Section V definitions and concepts of stability from
mathematics and control theory are reviewed to provide background information concerning stability of dynamic systems in
general and to establish theoretical connections.
The analytical definitions presented in Section V constitute
a key aspect of the report. They provide the mathematical underpinnings and bases for the definitions provided in the earlier
sections. These details are provided at the end of the report so
that interested readers can examine the finer points and assimilate the mathematical rigor.
0885-8950/04$20.00 © 2004 IEEE
Authorized licensed use limited to: Iowa State University. Downloaded on January 21, 2009 at 15:09 from IEEE Xplore. Restrictions apply.
1388
IEEE TRANSACTIONS ON POWER SYSTEMS, VOL. 19, NO. 2, MAY 2004
II. DEFINITION OF POWER SYSTEM STABILITY
In this section, we provide a formal definition of power
system stability. The intent is to provide a physically based
definition which, while conforming to definitions from system
theory, is easily understood and readily applied by power
system engineering practitioners.
A. Proposed Definition
• Power system stability is the ability of an electric power
system, for a given initial operating condition, to regain a
state of operating equilibrium after being subjected to a
physical disturbance, with most system variables bounded
so that practically the entire system remains intact.
B. Discussion and Elaboration
The definition applies to an interconnected power system as a
whole. Often, however, the stability of a particular generator or
group of generators is also of interest. A remote generator may
lose stability (synchronism) without cascading instability of the
main system. Similarly, stability of particular loads or load areas
may be of interest; motors may lose stability (run down and stall)
without cascading instability of the main system.
The power system is a highly nonlinear system that operates in a constantly changing environment; loads, generator outputs and key operating parameters change continually. When
subjected to a disturbance, the stability of the system depends
on the initial operating condition as well as the nature of the
disturbance.
Stability of an electric power system is thus a property of the
system motion around an equilibrium set, i.e., the initial operating condition. In an equilibrium set, the various opposing
forces that exist in the system are equal instantaneously (as in
the case of equilibrium points) or over a cycle (as in the case of
slow cyclical variations due to continuous small fluctuations in
loads or aperiodic attractors).
Power systems are subjected to a wide range of disturbances,
small and large. Small disturbances in the form of load changes
occur continually; the system must be able to adjust to the
changing conditions and operate satisfactorily. It must also
be able to survive numerous disturbances of a severe nature,
such as a short circuit on a transmission line or loss of a large
generator. A large disturbance may lead to structural changes
due to the isolation of the faulted elements.
At an equilibrium set, a power system may be stable for a
given (large) physical disturbance, and unstable for another. It
is impractical and uneconomical to design power systems to be
stable for every possible disturbance. The design contingencies
are selected on the basis they have a reasonably high probability
of occurrence. Hence, large-disturbance stability always refers
to a specified disturbance scenario. A stable equilibrium set thus
has a finite region of attraction; the larger the region, the more
robust the system with respect to large disturbances. The region
of attraction changes with the operating condition of the power
system.
The response of the power system to a disturbance may involve much of the equipment. For instance, a fault on a critical element followed by its isolation by protective relays will
cause variations in power flows, network bus voltages, and machine rotor speeds; the voltage variations will actuate both generator and transmission network voltage regulators; the generator speed variations will actuate prime mover governors; and
the voltage and frequency variations will affect the system loads
to varying degrees depending on their individual characteristics.
Further, devices used to protect individual equipment may respond to variations in system variables and cause tripping of the
equipment, thereby weakening the system and possibly leading
to system instability.
If following a disturbance the power system is stable, it will
reach a new equilibrium state with the system integrity preserved i.e., with practically all generators and loads connected
through a single contiguous transmission system. Some generators and loads may be disconnected by the isolation of faulted
elements or intentional tripping to preserve the continuity of operation of bulk of the system. Interconnected systems, for certain severe disturbances, may also be intentionally split into two
or more “islands” to preserve as much of the generation and
load as possible. The actions of automatic controls and possibly
human operators will eventually restore the system to normal
state. On the other hand, if the system is unstable, it will result
in a run-away or run-down situation; for example, a progressive increase in angular separation of generator rotors, or a progressive decrease in bus voltages. An unstable system condition
could lead to cascading outages and a shutdown of a major portion of the power system.
Power systems are continually experiencing fluctuations
of small magnitudes. However, for assessing stability when
subjected to a specified disturbance, it is usually valid to assume that the system is initially in a true steady-state operating
condition.
C. Conformance With System—Theoretic Definitions
In Section II-A, we have formulated the definition by considering a given operating condition and the system being subjected
to a physical disturbance. Under these conditions we require
the system to either regain a new state of operating equilibrium or return to the original operating condition (if no topological changes occurred in the system). These requirements are
directly correlated to the system-theoretic definition of asymptotic stability given in Section V-C-I. It should be recognized
here that this definition requires the equilibrium to be (a) stable
in the sense of Lyapunov, i.e., all initial conditions starting in
a small spherical neighborhood of radius result in the system
trajectory remaining in a cylinder of radius for all time
,
the initial time which corresponds to all of the system state varithe system trajecables being bounded, and (b) at time
tory approaches the equilibrium point which corresponds to the
equilibrium point being attractive. As a result, one observes that
the analytical definition directly correlates to the expected behavior in a physical system.
III. CLASSIFICATION OF POWER SYSTEM STABILITY
A typical modern power system is a high-order multivariable
process whose dynamic response is influenced by a wide array
of devices with different characteristics and response rates. Sta-
Authorized licensed use limited to: Iowa State University. Downloaded on January 21, 2009 at 15:09 from IEEE Xplore. Restrictions apply.
KUNDUR et al.: DEFINITION AND CLASSIFICATION OF POWER SYSTEM STABILITY
bility is a condition of equilibrium between opposing forces. Depending on the network topology, system operating condition
and the form of disturbance, different sets of opposing forces
may experience sustained imbalance leading to different forms
of instability. In this section, we provide a systematic basis for
classification of power system stability.
A. Need for Classification
Power system stability is essentially a single problem;
however, the various forms of instabilities that a power system
may undergo cannot be properly understood and effectively
dealt with by treating it as such. Because of high dimensionality and complexity of stability problems, it helps to make
simplifying assumptions to analyze specific types of problems
using an appropriate degree of detail of system representation
and appropriate analytical techniques. Analysis of stability,
including identifying key factors that contribute to instability
and devising methods of improving stable operation, is greatly
facilitated by classification of stability into appropriate categories [8]. Classification, therefore, is essential for meaningful
practical analysis and resolution of power system stability
problems. As discussed in Section V-C-I, such classification is
entirely justified theoretically by the concept of partial stability
[9]–[11].
B. Categories of Stability
The classification of power system stability proposed here is
based on the following considerations [8]:
• The physical nature of the resulting mode of instability as
indicated by the main system variable in which instability
can be observed.
• The size of the disturbance considered, which influences
the method of calculation and prediction of stability.
• The devices, processes, and the time span that must be
taken into consideration in order to assess stability.
Fig. 1 gives the overall picture of the power system stability
problem, identifying its categories and subcategories. The following are descriptions of the corresponding forms of stability
phenomena.
B.1 Rotor Angle Stability:
Rotor angle stability refers to the ability of synchronous machines of an interconnected power system to remain in synchronism after being subjected to a disturbance. It depends on the
ability to maintain/restore equilibrium between electromagnetic
torque and mechanical torque of each synchronous machine in
the system. Instability that may result occurs in the form of increasing angular swings of some generators leading to their loss
of synchronism with other generators.
The rotor angle stability problem involves the study of the
electromechanical oscillations inherent in power systems. A
fundamental factor in this problem is the manner in which
the power outputs of synchronous machines vary as their
rotor angles change. Under steady-state conditions, there is
1389
equilibrium between the input mechanical torque and the
output electromagnetic torque of each generator, and the speed
remains constant. If the system is perturbed, this equilibrium
is upset, resulting in acceleration or deceleration of the rotors
of the machines according to the laws of motion of a rotating
body. If one generator temporarily runs faster than another, the
angular position of its rotor relative to that of the slower machine will advance. The resulting angular difference transfers
part of the load from the slow machine to the fast machine,
depending on the power-angle relationship. This tends to
reduce the speed difference and hence the angular separation.
The power-angle relationship is highly nonlinear. Beyond a
certain limit, an increase in angular separation is accompanied
by a decrease in power transfer such that the angular separation
is increased further. Instability results if the system cannot
absorb the kinetic energy corresponding to these rotor speed
differences. For any given situation, the stability of the system
depends on whether or not the deviations in angular positions
of the rotors result in sufficient restoring torques [8]. Loss of
synchronism can occur between one machine and the rest of
the system, or between groups of machines, with synchronism
maintained within each group after separating from each other.
The change in electromagnetic torque of a synchronous
machine following a perturbation can be resolved into two
components:
• Synchronizing torque component, in phase with rotor
angle deviation.
• Damping torque component, in phase with the speed deviation.
System stability depends on the existence of both components
of torque for each of the synchronous machines. Lack of sufficient synchronizing torque results in aperiodic or nonoscillatory
instability, whereas lack of damping torque results in oscillatory
instability.
For convenience in analysis and for gaining useful insight into
the nature of stability problems, it is useful to characterize rotor
angle stability in terms of the following two subcategories:
• Small-disturbance (or small-signal) rotor angle stability
is concerned with the ability of the power system to maintain synchronism under small disturbances. The disturbances are considered to be sufficiently small that linearization of system equations is permissible for purposes
of analysis [8], [12], [13].
- Small-disturbance stability depends on the initial operating state of the system. Instability that may result
can be of two forms: i) increase in rotor angle through
a nonoscillatory or aperiodic mode due to lack of synchronizing torque, or ii) rotor oscillations of increasing
amplitude due to lack of sufficient damping torque.
- In today’s power systems, small-disturbance rotor
angle stability problem is usually associated with
insufficient damping of oscillations. The aperiodic
instability problem has been largely eliminated by use
of continuously acting generator voltage regulators;
however, this problem can still occur when generators
operate with constant excitation when subjected to the
actions of excitation limiters (field current limiters).
Authorized licensed use limited to: Iowa State University. Downloaded on January 21, 2009 at 15:09 from IEEE Xplore. Restrictions apply.
1390
IEEE TRANSACTIONS ON POWER SYSTEMS, VOL. 19, NO. 2, MAY 2004
Fig. 1.
Classification of power system stability.
- Small-disturbance rotor angle stability problems may
be either local or global in nature. Local problems
involve a small part of the power system, and are usually associated with rotor angle oscillations of a single
power plant against the rest of the power system. Such
oscillations are called local plant mode oscillations.
Stability (damping) of these oscillations depends on
the strength of the transmission system as seen by the
power plant, generator excitation control systems and
plant output [8].
- Global problems are caused by interactions among
large groups of generators and have widespread effects.
They involve oscillations of a group of generators in one
area swinging against a group of generators in another
area. Such oscillations are called interarea mode oscillations. Their characteristics are very complex and significantly differ from those of local plant mode oscillations. Load characteristics, in particular, have a major
effect on the stability of interarea modes [8].
- The time frame of interest in small-disturbance stability studies is on the order of 10 to 20 seconds following a disturbance.
• Large-disturbance rotor angle stability or transient stability, as it is commonly referred to, is concerned with the
ability of the power system to maintain synchronism when
subjected to a severe disturbance, such as a short circuit
on a transmission line. The resulting system response involves large excursions of generator rotor angles and is
influenced by the nonlinear power-angle relationship.
- Transient stability depends on both the initial
operating state of the system and the severity of the disturbance. Instability is usually in the form of aperiodic
angular separation due to insufficient synchronizing
torque, manifesting as first swing instability. However,
in large power systems, transient instability may not
always occur as first swing instability associated with
a single mode; it could be a result of superposition of
a slow interarea swing mode and a local-plant swing
mode causing a large excursion of rotor angle beyond
the first swing [8]. It could also be a result of nonlinear
effects affecting a single mode causing instability
beyond the first swing.
- The time frame of interest in transient stability studies
is usually 3 to 5 seconds following the disturbance. It
may extend to 10–20 seconds for very large systems
with dominant inter-area swings.
As identified in Fig. 1, small-disturbance rotor angle stability
as well as transient stability are categorized as short term
phenomena.
The term dynamic stability also appears in the literature as
a class of rotor angle stability. However, it has been used to
denote different phenomena by different authors. In the North
American literature, it has been used mostly to denote small-disturbance stability in the presence of automatic controls (particularly, the generation excitation controls) as distinct from the
classical “steady-state stability” with no generator controls [7],
[8]. In the European literature, it has been used to denote transient stability. Since much confusion has resulted from the use
of the term dynamic stability, we recommend against its usage,
as did the previous IEEE and CIGRE Task Forces [6], [7].
B.2 Voltage Stability:
Voltage stability refers to the ability of a power system to maintain steady voltages at all buses in the system after being subjected to a disturbance from a given initial operating condition.
It depends on the ability to maintain/restore equilibrium between load demand and load supply from the power system. Instability that may result occurs in the form of a progressive fall
or rise of voltages of some buses. A possible outcome of voltage
instability is loss of load in an area, or tripping of transmission lines and other elements by their protective systems leading
Authorized licensed use limited to: Iowa State University. Downloaded on January 21, 2009 at 15:09 from IEEE Xplore. Restrictions apply.
KUNDUR et al.: DEFINITION AND CLASSIFICATION OF POWER SYSTEM STABILITY
to cascading outages. Loss of synchronism of some generators
may result from these outages or from operating conditions that
violate field current limit [14].
Progressive drop in bus voltages can also be associated with
rotor angle instability. For example, the loss of synchronism
of machines as rotor angles between two groups of machines
approach 180 causes rapid drop in voltages at intermediate
points in the network close to the electrical center [8]. Normally, protective systems operate to separate the two groups
of machines and the voltages recover to levels depending on
the post-separation conditions. If, however, the system is not
so separated, the voltages near the electrical center rapidly oscillate between high and low values as a result of repeated “pole
slips” between the two groups of machines. In contrast, the type
of sustained fall of voltage that is related to voltage instability
involves loads and may occur where rotor angle stability is not
an issue.
The term voltage collapse is also often used. It is the process
by which the sequence of events accompanying voltage instability leads to a blackout or abnormally low voltages in a significant part of the power system [8], [15], and [16]. Stable
(steady) operation at low voltage may continue after transformer
tap changers reach their boost limit, with intentional and/or unintentional tripping of some load. Remaining load tends to be
voltage sensitive, and the connected demand at normal voltage
is not met.
The driving force for voltage instability is usually the loads;
in response to a disturbance, power consumed by the loads tends
to be restored by the action of motor slip adjustment, distribution
voltage regulators, tap-changing transformers, and thermostats.
Restored loads increase the stress on the high voltage network
by increasing the reactive power consumption and causing further voltage reduction. A run-down situation causing voltage instability occurs when load dynamics attempt to restore power
consumption beyond the capability of the transmission network
and the connected generation [8], [14]–[18].
A major factor contributing to voltage instability is the
voltage drop that occurs when active and reactive power flow
through inductive reactances of the transmission network; this
limits the capability of the transmission network for power
transfer and voltage support. The power transfer and voltage
support are further limited when some of the generators hit
their field or armature current time-overload capability limits.
Voltage stability is threatened when a disturbance increases the
reactive power demand beyond the sustainable capacity of the
available reactive power resources.
While the most common form of voltage instability is the
progressive drop of bus voltages, the risk of overvoltage instability also exists and has been experienced at least on one system
[19]. It is caused by a capacitive behavior of the network (EHV
transmission lines operating below surge impedance loading) as
well as by underexcitation limiters preventing generators and/or
synchronous compensators from absorbing the excess reactive
power. In this case, the instability is associated with the inability of the combined generation and transmission system to
operate below some load level. In their attempt to restore this
load power, transformer tap changers cause long-term voltage
instability.
1391
Voltage stability problems may also be experienced at the
terminals of HVDC links used for either long distance or
back-to-back applications [20], [21]. They are usually associated with HVDC links connected to weak ac systems and
may occur at rectifier or inverter stations, and are associated
with the unfavorable reactive power “load” characteristics
of the converters. The HVDC link control strategies have a
very significant influence on such problems, since the active
and reactive power at the ac/dc junction are determined by
the controls. If the resulting loading on the ac transmission
stresses it beyond its capability, voltage instability occurs. Such
a phenomenon is relatively fast with the time frame of interest
being in the order of one second or less. Voltage instability
may also be associated with converter transformer tap-changer
controls, which is a considerably slower phenomenon [21].
Recent developments in HVDC technology (voltage source
converters and capacitor commutated converters) have significantly increased the limits for stable operation of HVDC
links in weak systems as compared with the limits for line
commutated converters.
One form of voltage stability problem that results in uncontrolled overvoltages is the self-excitation of synchronous machines. This can arise if the capacitive load of a synchronous
machine is too large. Examples of excessive capacitive loads
that can initiate self-excitation are open ended high voltage lines
and shunt capacitors and filter banks from HVDC stations [22].
The overvoltages that result when generator load changes to capacitive are characterized by an instantaneous rise at the instant
of change followed by a more gradual rise. This latter rise depends on the relation between the capacitive load component
and machine reactances together with the excitation system of
the synchronous machine. Negative field current capability of
the exciter is a feature that has a positive influence on the limits
for self-excitation.
As in the case of rotor angle stability, it is useful to classify
voltage stability into the following subcategories:
• Large-disturbance voltage stability refers to the system’s
ability to maintain steady voltages following large disturbances such as system faults, loss of generation, or
circuit contingencies. This ability is determined by the
system and load characteristics, and the interactions of
both continuous and discrete controls and protections. Determination of large-disturbance voltage stability requires
the examination of the nonlinear response of the power
system over a period of time sufficient to capture the performance and interactions of such devices as motors, underload transformer tap changers, and generator field-current limiters. The study period of interest may extend from
a few seconds to tens of minutes.
• Small-disturbance voltage stability refers to the system’s
ability to maintain steady voltages when subjected to small
perturbations such as incremental changes in system load.
This form of stability is influenced by the characteristics of
loads, continuous controls, and discrete controls at a given
instant of time. This concept is useful in determining, at
any instant, how the system voltages will respond to small
system changes. With appropriate assumptions, system
equations can be linearized for analysis thereby allowing
Authorized licensed use limited to: Iowa State University. Downloaded on January 21, 2009 at 15:09 from IEEE Xplore. Restrictions apply.
1392
IEEE TRANSACTIONS ON POWER SYSTEMS, VOL. 19, NO. 2, MAY 2004
computation of valuable sensitivity information useful in
identifying factors influencing stability. This linearization,
however, cannot account for nonlinear effects such as
tap changer controls (deadbands, discrete tap steps, and
time delays). Therefore, a combination of linear and
nonlinear analyzes is used in a complementary manner
[23], [24].
As noted above, the time frame of interest for voltage stability problems may vary from a few seconds to tens of minutes. Therefore, voltage stability may be either a short-term or
a long-term phenomenon as identified in Figure 1.
• Short-term voltage stability involves dynamics of fast
acting load components such as induction motors,
electronically controlled loads, and HVDC converters.
The study period of interest is in the order of several
seconds, and analysis requires solution of appropriate
system differential equations; this is similar to analysis
of rotor angle stability. Dynamic modeling of loads is
often essential. In contrast to angle stability, short circuits
near loads are important. It is recommended that the term
transient voltage stability not be used.
• Long-term voltage stability involves slower acting equipment such as tap-changing transformers, thermostatically
controlled loads, and generator current limiters. The study
period of interest may extend to several or many minutes,
and long-term simulations are required for analysis of
system dynamic performance [17], [23], [25]. Stability
is usually determined by the resulting outage of equipment, rather than the severity of the initial disturbance.
Instability is due to the loss of long-term equilibrium
(e.g., when loads try to restore their power beyond the
capability of the transmission network and connected
generation), post-disturbance steady-state operating point
being small-disturbance unstable, or a lack of attraction
toward the stable post-disturbance equilibrium (e.g., when
a remedial action is applied too late) [14], [15]. The
disturbance could also be a sustained load buildup (e.g.,
morning load increase). In many cases, static analysis
[23], [24], [26], [27] can be used to estimate stability
margins, identify factors influencing stability, and screen
a wide range of system conditions and a large number
of scenarios. Where timing of control actions is important, this should be complemented by quasi-steady-state
time-domain simulations [14], [17].
B.3 Basis for Distinction between Voltage and
Rotor Angle Stability:
It is important to recognize that the distinction between rotor
angle stability and voltage stability is not based on weak
coupling between variations in active power/angle and reactive power/voltage magnitude. In fact, coupling is strong for
stressed conditions and both rotor angle stability and voltage
stability are affected by pre-disturbance active power as well
as reactive power flows. Instead, the distinction is based on
the specific set of opposing forces that experience sustained
imbalance and the principal system variable in which the
consequent instability is apparent.
B.4 Frequency Stability:
Frequency stability refers to the ability of a power system to
maintain steady frequency following a severe system upset resulting in a significant imbalance between generation and load.
It depends on the ability to maintain/restore equilibrium between system generation and load, with minimum unintentional
loss of load. Instability that may result occurs in the form of sustained frequency swings leading to tripping of generating units
and/or loads.
Severe system upsets generally result in large excursions of
frequency, power flows, voltage, and other system variables,
thereby invoking the actions of processes, controls, and protections that are not modeled in conventional transient stability
or voltage stability studies. These processes may be very slow,
such as boiler dynamics, or only triggered for extreme system
conditions, such as volts/Hertz protection tripping generators.
In large interconnected power systems, this type of situation is
most commonly associated with conditions following splitting
of systems into islands. Stability in this case is a question of
whether or not each island will reach a state of operating equilibrium with minimal unintentional loss of load. It is determined
by the overall response of the island as evidenced by its mean
frequency, rather than relative motion of machines. Generally,
frequency stability problems are associated with inadequacies
in equipment responses, poor coordination of control and protection equipment, or insufficient generation reserve. Examples
of such problems are reported in references [28]–[31]. In isolated island systems, frequency stability could be of concern for
any disturbance causing a relatively significant loss of load or
generation [32].
During frequency excursions, the characteristic times of the
processes and devices that are activated will range from fraction
of seconds, corresponding to the response of devices such as
underfrequency load shedding and generator controls and protections, to several minutes, corresponding to the response of
devices such as prime mover energy supply systems and load
voltage regulators. Therefore, as identified in Fig. 1, frequency
stability may be a short-term phenomenon or a long-term phenomenon. An example of short-term frequency instability is
the formation of an undergenerated island with insufficient underfrequency load shedding such that frequency decays rapidly
causing blackout of the island within a few seconds [28]. On
the other hand, more complex situations in which frequency
instability is caused by steam turbine overspeed controls [29]
or boiler/reactor protection and controls are longer-term phenomena with the time frame of interest ranging from tens of
seconds to several minutes [30], [31], [33].
During frequency excursions, voltage magnitudes may
change significantly, especially for islanding conditions with
underfrequency load shedding that unloads the system. Voltage
magnitude changes, which may be higher in percentage than
frequency changes, affect the load-generation imbalance.
High voltage may cause undesirable generator tripping by
poorly designed or coordinated loss of excitation relays or
volts/Hertz relays. In an overloaded system, low voltage may
cause undesirable operation of impedance relays.
Authorized licensed use limited to: Iowa State University. Downloaded on January 21, 2009 at 15:09 from IEEE Xplore. Restrictions apply.
KUNDUR et al.: DEFINITION AND CLASSIFICATION OF POWER SYSTEM STABILITY
B.5 Comments on Classification:
We have classified power system stability for convenience in
identifying causes of instability, applying suitable analysis
tools, and developing corrective measures. In any given situation, however, any one form of instability may not occur in its
pure form. This is particularly true in highly stressed systems
and for cascading events; as systems fail one form of instability
may ultimately lead to another form. However, distinguishing
between different forms is important for understanding the underlying causes of the problem in order to develop appropriate
design and operating procedures.
While classification of power system stability is an effective and convenient means to deal with the complexities of the
problem, the overall stability of the system should always be
kept in mind. Solutions to stability problems of one category
should not be at the expense of another. It is essential to look at
all aspects of the stability phenomenon, and at each aspect from
more than one viewpoint.
IV. RELATIONSHIP BETWEEN RELIABILITY, SECURITY,
AND STABILITY
In this section, we discuss the relationship between the concepts of power system reliability, security, and stability. We will
also briefly describe how these terms have been defined and
used in practice.
A. Conceptual Relationship [34], [35]
Reliability of a power system refers to the probability of its
satisfactory operation over the long run. It denotes the ability to
supply adequate electric service on a nearly continuous basis,
with few interruptions over an extended time period.
Security of a power system refers to the degree of risk in
its ability to survive imminent disturbances (contingencies)
without interruption of customer service. It relates to robustness
of the system to imminent disturbances and, hence, depends
on the system operating condition as well as the contingent
probability of disturbances.
Stability of a power system, as discussed in Section II, refers
to the continuance of intact operation following a disturbance. It
depends on the operating condition and the nature of the physical disturbance.
The following are the essential differences among the three
aspects of power system performance:
1) Reliability is the overall objective in power system
design and operation. To be reliable, the power system
must be secure most of the time. To be secure, the system
must be stable but must also be secure against other
contingencies that would not be classified as stability
problems e.g., damage to equipment such as an explosive
failure of a cable, fall of transmission towers due to ice
loading or sabotage. As well, a system may be stable
following a contingency, yet insecure due to post-fault
system conditions resulting in equipment overloads or
voltage violations.
2) System security may be further distinguished from
stability in terms of the resulting consequences. For
1393
example, two systems may both be stable with equal
stability margins, but one may be relatively more secure
because the consequences of instability are less severe.
3) Security and stability are time-varying attributes which
can be judged by studying the performance of the power
system under a particular set of conditions. Reliability, on
the other hand, is a function of the time-average performance of the power system; it can only be judged by consideration of the system’s behavior over an appreciable
period of time.
B. NERC Definition of Reliability [36]
NERC (North American Electric Reliability Council) defines
power system reliability as follows.
• Reliability, in a bulk power electric system, is the degree
to which the performance of the elements of that system
results in power being delivered to consumers within accepted standards and in the amount desired. The degree
of reliability may be measured by the frequency, duration,
and magnitude of adverse effects on consumer service.
Reliability can be addressed by considering two basic functional aspects of the power systems:
Adequacy—the ability of the power system to supply the aggregate electric power and energy requirements of the customer
at all times, taking into account scheduled and unscheduled outages of system components.
Security—the ability of the power system to withstand sudden
disturbances such as electric short circuits or nonanticipated loss
of system components.
The above definitions also appear in several IEEE and CIGRE
Working Group/Task Force documents [37], [38].
Other alternative forms of definition of power system security
have been proposed in the literature. For example, in reference
[39], security is defined in terms of satisfying a set of inequality
constraints over a subset of the possible disturbances called the
“next contingency set.”
C. Analysis of Power System Security
The analysis of security relates to the determination of the robustness of the power system relative to imminent disturbances.
There are two important components of security analysis. For a
power system subjected to changes (small or large), it is important that when the changes are completed, the system settles to
new operating conditions such that no physical constraints are
violated. This implies that, in addition to the next operating conditions being acceptable, the system must survive the transition
to these conditions.
The above characterization of system security clearly highlights two aspects of its analysis:
• Static security analysis—This involves steady-state analysis of post-disturbance system conditions to verify that
no equipment ratings and voltage constraints are violated.
• Dynamic security analysis—This involves examining
different categories of system stability described in
Section III.
Stability analysis is thus an integral component of system security and reliability assessment.
Authorized licensed use limited to: Iowa State University. Downloaded on January 21, 2009 at 15:09 from IEEE Xplore. Restrictions apply.
1394
IEEE TRANSACTIONS ON POWER SYSTEMS, VOL. 19, NO. 2, MAY 2004
The general industry practice for security assessment has
been to use a deterministic approach. The power system is
designed and operated to withstand a set of contingencies
referred to as “normal contingencies” selected on the basis that
they have a significant likelihood of occurrence. In practice,
they are usually defined as the loss of any single element in a
power system either spontaneously or preceded by a single-,
double-, or three-phase fault. This is usually referred to as
criterion because it examines the behavior of an
the
-component grid following the loss of any one of its major
components. In addition, loss of load or cascading outages
may not be allowed for multiple-related outages such as loss of
a double-circuit line. Consideration may be given to extreme
contingencies that exceed in severity the normal design contingencies. Emergency controls, such as generation tripping, load
shedding, and controlled islanding, may be used to cope with
such events and prevent widespread blackouts.
The deterministic approach has served the industry reasonably well in the past—it has resulted in high security levels and
the study effort is minimized. Its main limitation, however, is
that it treats all security-limiting scenarios as having the same
risk. It also does not give adequate consideration as to how likely
or unlikely various contingencies are.
In today’s utility environment, with a diversity of participants
with different business interests, the deterministic approach may
not be acceptable. There is a need to account for the probabilistic nature of system conditions and events, and to quantify and manage risk. The trend will be to expand the use of
risk-based security assessment. In this approach, the probability
of the system becoming unstable and its consequences are examined, and the degree of exposure to system failure is estimated. This approach is computationally intensive but is possible with today’s computing and analysis tools.
V. SYSTEM-THEORETIC FOUNDATIONS OF POWER
SYSTEM STABILITY
A. Preliminaries
In this section, we address fundamental issues related to definitions of power system stability from a system-theoretic viewpoint. We assume that the model of a power system is given
in the form of explicit first-order differential equations (i.e.,
a state-space description). While this is quite common in the
theory of dynamical systems, it may not always be entirely natural for physical systems such as power systems. First-principle
models that are typically used to describe power systems are
seldom in this form, and transformations required to bring them
to explicit first-order form may, in general, introduce spurious
solutions [40].
More importantly, there often exist algebraic (implicit) equations that constrain various quantities, and a set of differential-algebraic equations (DAE) is often used in simulations of
power system transients. The algebraic part often arises from a
singular perturbation-type reasoning that uses time separation
between subsets of variables to postulate that the fast variables
have already reached their steady state on the time horizon of
interest [41], [42].
Proving the existence of solutions of DAE is a very challenging problem in general. While local results can be derived
from the implicit function theorem that specifies rank conditions
for the Jacobian of the algebraic part, the non-local results are
much harder to obtain. One general approach to non-local study
of stability of DAE systems that is based on differential geometry is presented in [43] (for sufficient conditions see [44]). The
surfaces on which rank conditions for the Jacobian of the algebraic part do not hold are commonly denoted as impasse surfaces, and in the analysis of models of power systems, it is typically assumed that equilibrium sets of interest in stability analysis are disjoint from such surfaces [41], [45].
An often useful approximation of the fast dynamics is based
on the concept of dynamic (time-varying) phasors [46] and
dynamic symmetrical components [47]. It is also typically
assumed that distributed nature of some elements of a power
system (e.g., transmission lines) can be approximated with
lumped parameter models without a major loss of model fidelity. This is mostly dictated by the fundamental intractability
of models that include partial differential equations, and by
satisfactory behavior of lumped parameter models (when
evaluated on the level of single element—e.g., the use of
multiple “ ” section models for a long transmission line).
Some qualitative aspects of fault propagation in spatially
extended power systems can, however, be studied effectively
with distributed models [48].
Power systems are also an example of constrained dynamical
systems, as their state trajectories are restricted to a particular
subset in the state space (phase space in the language of dynamical systems) denoted as the feasible (or technically viable,
or permitted) operating region [45]. The trajectories that exit
this desired region may either lead to structural changes (e.g.,
breaker tripping in a power system), or lead to unsafe operation.
This type of consideration will introduce restricted stability regions in power system stability analysis.
Several additional issues are raised by the fact that the
power system interacts with its (typically unmodeled) environment, making the power system model nonautonomous (or
time-varying). Examples include load variations and network
topology changes due to switching in substations. Additional
interactions with the environment include disturbances whose
physical description may include outages of system elements,
while a mathematical description may involve variations in the
system order, or the number of variables of interest.
Finally, a power system is a controlled (or forced) system
with numerous feedback loops, and it is necessary to include the
effects of control inputs (including their saturation), especially
on longer time horizons.
The outlined modeling problems are typically addressed in a
power system analysis framework in the following way:
1) The problem of defining stability for general nonautonomous systems is very challenging even in the
theoretical realm [40], and one possible approach is to
say that a system to which the environment delivers
square-integrable signals as inputs over a time interval
is stable if variables of interest (such as outputs) are
also square integrable. In a more general setup, we can
consider signals truncated in time, and denote the system
Authorized licensed use limited to: Iowa State University. Downloaded on January 21, 2009 at 15:09 from IEEE Xplore. Restrictions apply.
KUNDUR et al.: DEFINITION AND CLASSIFICATION OF POWER SYSTEM STABILITY
as well-posed if it maps square integrable truncated
signals into signals with the same property. In a power
system setting, one typically assumes that the variables
at the interface with the environment are known (or predictable)—e.g., that mechanical inputs to all generators
are constant, or that they vary according to the known
response of turbine regulators.
2) The disturbances of interest will fall into two broad
categories—event-type (typically described as outages of
specific pieces of equipment) and norm-type (described
by their size e.g., in terms of various norms of signals);
we will return to this issue shortly. We also observe that
in cases when event-type (e.g., switching) disturbances
occur repeatedly, a proper analysis framework is that of
hybrid systems (for a recent review, see [49]); event-type
disturbances may also be initiated by human operators.
Our focus is on time horizons of the order of seconds
to minutes; on a longer time scale, the effects of market
structures may become prominent. In that case, the
relevant notion of stability needs re-examination; some
leads about systems with distributed decision making
may be found in [50].
3) Given our emphasis on stability analysis, we will assume
that the actions of all controllers are fully predictable in
terms of known system quantities (states), or as functions
of time; the dual problem of designing stabilizing controls
for nonlinear systems is very challenging, see for example
[8], [51].
A typical power system stability study consists of the following steps:
1) Make modeling assumptions and formulate a mathematical model appropriate for the time-scales and phenomena
under study;
2) Select an appropriate stability definition;
3) Analyze and/or simulate to determine stability, typically
using a scenario of events;
4) Review results in light of assumptions, compare with
the engineering experience (“reality”), and repeat if
necessary.
Before considering specifics about power system stability, we
need to assess the required computational effort. In the case of
linear system models, the stability question is decidable, and
can be answered efficiently in polynomial time [52]. In the case
of nonlinear systems, the available algorithms are inherently inefficient. For example, a related problem of whether all trajectories of a system with a single scalar nonlinearity converge to
the origin turns out to be very computationally intensive (i.e.,
NP-hard), and it is unclear if it is decidable at all [52]. Given the
large size of power systems and the need to consider event-type
perturbations that will inevitably lead to nonlinear models, it is
clear that the task of determining stability of a power system will
be a challenging one. It turns out, however, that our main tools in
reducing the computational complexity will be our ability (and
willingness) to utilize approximations, and the particular nature
of event-type disturbances that we are analyzing.
We also want to point out that a possible shift in emphasis regarding various phenomena in power systems (e.g., hybrid aspects) would necessarily entail a reassessment of notions of sta-
1395
bility. For a recent review on notions of stability in various types
of systems (including infinite dimensional ones), see [53].
B. A Scenario for Stability Analysis
We consider the system
where is the state vector (a function of time, but we omit explicitly writing the time argument ), is its derivative, is sufficiently differentiable and its domain includes the origin. The
is
system described above is said to be autonomous if
independent of and is said to be nonautonomous otherwise.
A typical scenario for power system stability analysis
involves three distinct steps.
1) The system is initially operating in a pre-disturbance
(e.g., an equilibrium point or perhaps
equilibrium set
even a benign limit cycle in the state space); in that set,
various driving terms (forces) affecting system variables
are balanced (either instantaneously, or over a time
interval). We use the notion of an equilibrium set to
denote equilibrium points, limit sets and more complicated structures like aperiodic attractors (which may be
possible in realistic models of power systems). However,
in the vast majority of cases of practical interest today,
the equilibrium points are the sets of interest.
In general, an equilibrium set, or an attractor, is a
set of trajectories in the phase space to which all neighboring trajectories converge. Attractors therefore describe
the long-term behavior of a dynamical system. A Point attractor, or an equilibrium point, is an attractor consisting
of a single point in the phase space. A Limit cycle attractor, on the other hand, corresponds to closed curves
in phase space; limit cycles imply periodic behavior. A
chaotic (or aperiodic, or strange) attractor corresponds
to a equilibrium set where system trajectories never converge to a point or a closed curve, but remain within the
same region of phase space. Unlike limit cycles, strange
attractors are non-periodic, and trajectories in such systems are very sensitive to the initial conditions.
2) Next, a disturbance acts on the system. An event-type
(or incident-type) disturbance is characterized by a specific fault scenario (e.g., short circuit somewhere in the
transmission network followed by a line disconnection including the duration of the event—“fault clearing time”),
while norm-type (described by their size in terms of various norms of signals—e.g., load variations) disturbances
are described by their size (norm, or signal intensity). A
problem of some analytical interest is determining the
maximum permissible duration of the fault (the so called
“critical clearing time”) for which the subsequent system
response remains stable. This portion of stability analysis
requires the knowledge of actions of protective relaying.
3) After an event-type disturbance, the system dynamics is
studied with respect to a known post-disturbance equi(which may be distinct from
). The
librium set
system initial condition belongs to a (known) starting set
, and we want to characterize the system motion with
Authorized licensed use limited to: Iowa State University. Downloaded on January 21, 2009 at 15:09 from IEEE Xplore. Restrictions apply.
1396
IEEE TRANSACTIONS ON POWER SYSTEMS, VOL. 19, NO. 2, MAY 2004
respect to
i.e., if the system trajectory will remain in(which includes
). In
side the technically viable set
the case of norm-type disturbances, very often we have
. If the system response turns out to be stable
(a precise definition will follow shortly), it is said that
(and sometimes
as well) are stable. A detected instability (during which system motion crosses the boundary
—e.g., causing line tripof the technically viable set
ping or a partial load shedding) may lead to a new stability
study for a new (reduced) system with new starting and
viable sets, and possibly with different modeling assumptions (or several such studies, if a system gets partitioned
into several disconnected parts).
The stability analysis of power systems is in general nonlocal, as various equilibrium sets may get involved. In the case of
event-type disturbances, the perturbations of interest are specias well), and
fied deterministically (the same may apply to
that are releit is assumed that the analyst has determined all
and the disturbance. In the case of norm-type
vant for a given
perturbations, the uncertainty structure is different—the perturbation is characterized by size (in the case of the so-called
small-disturbance or small-signal analysis this is done implicitly, so that linearized analysis remains valid), and the same
equilibrium set typically characterizes the system before and
after the disturbance. Note, however, that norm-type perturbations could in principle be used in large-signal analyses as well.
We propose next a formulation of power system stability that
will allow us to explore salient features of general stability concepts from system theory.
• “An equilibrium set of a power system is stable if, when
the initial state is in the given starting set, the system motion converges to the equilibrium set, and operating constraints are satisfied for all relevant variables along the
entire trajectory.”
The operating constraints are of inequality (and equality)
type, and pertain to individual variables and their collections.
For example, system connectedness is a collective feature,
as it implies that there exist paths (in graph-theoretic terms)
from any given bus to all other buses in the network. Note also
that some of the operating constraints (e.g., voltage levels) are
inherently soft i.e., the power system analyst may be interested
in stability characterization with and without these constraints.
Note that we assume that our model is accurate within
in the sense that there are no further system changes (e.g.,
relay-initiated line tripping) until the trajectory crosses the
.
boundary
C. Stability Definitions From System Theory
In this section, we provide detailed analytical definitions of
several types of stability including Lyapunov stability, inputoutput stability, stability of linear systems, and partial stability.
Of these various types, the Lyapunov stability definitions related
to stability and asymptotic stability are the ones most applicable
to power system nonlinear behavior under large disturbances.
The definition of stability related to linear systems finds wide
use in small-signal stability analysis of power systems. The con-
Fig. 2. Illustration of the definition of stability [57].
cept of partial stability is useful in the classification of power
system stability into different categories.
C.1 Lyapunov Stability:
The definitions collected here mostly follow the presentation in
[54]; we present definitions for cases that are typical for power
system models (e.g., assuming differentiability of functions involved), and not necessarily in the most general context. We
will concentrate on the study of stability of equilibrium points;
a study of more intricate equilibrium sets, like periodic orbits,
can often be reduced to the study of equilibrium points of an associated system whose states are deviations of the states of the
original system from the periodic orbit; another possibility is to
study periodic orbits via sampled states and Poincare maps [55].
We again consider the nonautonomous system:
(1)
where is the state vector, is its derivative, is sufficiently
differentiable and its domain includes the origin. Note that the
forcing (control input) term is not included i.e., we do not write
. In stability analysis in the Lyapunov framework that
is not a limitation, since all control inputs are assumed to be
known functions of time and/or known functions of states .
For technical reasons, we will assume that the origin is an equi,
). An equiliblibrium point (meaning that
rium at the origin could be a translation of a nonzero equilibrium
after a suitable coordinate transformation [54, p. 132]
of (1) is:
The equilibrium point
• stable if, for each
, there is
such
that:
(2)
Note that in (2) any norm can be used due to topological
equivalence of all norms. In Figure 2, we depict the behavior
of trajectories in the vicinity of a stable equilibrium for the case
(a two-dimensional system in the space of real variables). By choosing the initial conditions in a sufficiently small
spherical neighborhood of radius , we can force the trajectory
of the system for all time
to lie entirely in a given cylinder
of radius .
• uniformly stable if, for each
, there is
,
independent of , such that (2) is satisfied;
• unstable if not stable;
• asymptotically stable if it is stable and in addition there is
such that:
Authorized licensed use limited to: Iowa State University. Downloaded on January 21, 2009 at 15:09 from IEEE Xplore. Restrictions apply.
KUNDUR et al.: DEFINITION AND CLASSIFICATION OF POWER SYSTEM STABILITY
Fig. 3.
1397
Illustration of the definition of asymptotic stability [57].
Fig. 5. Illustration of the definition of exponential stability [58].
Fig. 4.
Illustration of the definition of uniform asymptotic stability [58].
It is important to note that the definition of asymptotic stability combines the aspect of stability as well as attractivity of
the equilibrium. This is a stricter requirement of the system behavior to eventually return to the equilibrium point. This concept
is pictorially presented in Figure 3.
• uniformly asymptotically stable if it is uniformly stable
and there is
, independent of , such that for all
,
as
, uniformly in and
; that is, for each
, there is
such that
In Figure 4, we depict the property of uniform asymptotic
stability pictorially. By choosing the initial operating points in
, we can
a sufficiently small spherical neighborhood at
force the trajectory of the solution to lie inside a given cylinder
.
for all
• globally uniformly asymptotically stable if it is uniformly
stable, and, for each pair of positive numbers and ,
such that:
there is
• exponentially stable if there are
that:
,
,
such
In Figure 5 the behavior of a solution in the vicinity of an
exponentially stable equilibrium point is shown.
• globally exponentially stable if the exponential stability
condition is satisfied for any initial state.
These definitions form the foundation of the Lyapunov approach to system stability, and can be most naturally checked
for a specific system via so called Lyapunov functions. Qualitatively speaking, i.e., disregarding subtleties due to the nonautonomous characteristics of the system, we are to construct a
smooth positive definite “energy” function whose time derivative (along trajectories of (1)) is negative definite. While unfortunately there is no systematic method to generate such functions (some leads for the case of simple power systems are given
in [56], [57]), the so called converse Lyapunov theorems establish the existence of such functions if the system is stable in a
certain sense.
In power systems we are interested in, the region of attraction
of a given equilibrium set
, namely the set of points
in the state space with the property that all trajectories initiated
. If the equiat the points will converge to the equilibrium set
librium set is a point that is asymptotically stable, then it can be
shown that the region of attraction has nice analytical properties
(it is an open and connected set, and its boundary is formed by
system trajectories). In the case of large scale power systems, we
are naturally interested in the effects of approximations and idealizations that are necessary because of system size. Even if the
nominal system has a stable equilibrium at the origin, this may
not be the case for the actual perturbed system, which is not entirely known to the analyst. We cannot necessarily expect that
the solution of the perturbed system will approach the origin,
is
but could if the solution is ultimately bounded i.e.,
bounded by a fixed constant, given that the initial condition is
in a ball of a fixed radius, and for sufficiently large time . Characterization of stability in this case requires knowledge of the
size of the perturbation term, and of a Lyapunov function for the
nominal (non-perturbed) system. A related notion of practical
stability is motivated by the idea that a system may be “considered stable if the deviations of motions from the equilibrium
remain within certain bounds determined by the physical situation, in case the initial values and the perturbation are bounded
by suitable constants” [59]. One does not require a more narrow
can be
interpretation that the deviation from the origin of
made arbitrarily small by a suitable choice of the constants, as is
the case with total stability. Roughly speaking, for practical stability, we allow that the system will move away from the origin
even for small perturbations, and we cannot make that motion
arbitrarily small by reducing the model perturbation term.
Another concept of interest in power systems is that of partial stability [9]–[11], introduced already by Lyapunov himself.
The basic idea is to relax the condition for stability from one that
Authorized licensed use limited to: Iowa State University. Downloaded on January 21, 2009 at 15:09 from IEEE Xplore. Restrictions apply.
1398
IEEE TRANSACTIONS ON POWER SYSTEMS, VOL. 19, NO. 2, MAY 2004
requires stable behavior from all variables (because of the properties of the norm used in (2) and elsewhere) to one that requires
such behavior from only some of the variables. This formulation
is natural in some engineered systems [9], and leads to substantial simplifications in others (e.g., in some adaptive systems). It
has been used in the context of power system stability as well
[60].
A power system is often modeled as an interconnection of
lower-order subsystems, and we may be interested in a hierarchical (two-level) approach to stability determination [61]. At
the first step, we analyze the stability of each subsystem separately (i.e., while ignoring the interconnections). In the second
step, we combine the results of the first step with information
about the interconnections to analyze the stability of the overall
system. In a Lyapunov framework, this results in the study of
composite Lyapunov functions. An important qualitative result
is that if the isolated subsystems are sufficiently stable, compared to the strength of the interconnections, then the overall
system is uniformly asymptotically stable at the origin.
C.2 Input/Output Stability:
This approach considers the system description of the form:
(3)
where is an operator (nonlinear in general) that specifies the
-dimensional output vector in terms of the -dimensional
input . The input belongs to a normed linear space of vector
signals
—e.g., extended bounded or square integrable sigof such signals (set to zero
nals, meaning that all truncations
) are bounded or square integrable (this allows inclufor
sion of “growing” signals like ramps etc. that are of interest in
stability analysis).
• Definition: A continuous function
is said to belong to class if it is strictly increasing and
.
• Definition: A continuous function
is said to belong to class
if, for each fixed ,
belongs to class with respect to
the mapping
and, for each fixed , the mapping
is decreasing
as
.
with respect to and
is stable if there exists a
• A mapping :
defined on
, and a non-negaclass- function
tive constant such that:
(4)
for all
and
.
It is finite-gain stable if there exist non-negative constants
and such that:
(5)
for all
and
.
is the space of uniformly bounded signals,
Note that if
then this definition yields the familiar notion of bounded-input,
bounded-output stability. The above definitions exclude systems
for which inequalities (4) and (5) are defined only for a subset
of the input space; this is allowed in the notion of small-signal
stability, where the norm of the input signals is constrained.
Let us consider a nonautonomous system with input:
(6)
Note the shift in analytical framework, as in input/output stability inputs are not assumed to be known functions of time, but
assumed to be in a known class typically described by a norm.
A system (6) is said to be locally input-to-state stable if there
function
, a class- function
, and
exists a classpositive constants and such that for any initial state
with
and any input
with
, the solution
exists and satisfies:
(7)
for all
.
It is said to be input-to-state stable if the local input-to-state
property holds for the entire input and output spaces, and
and any
inequality (7) is satisfied for any initial state
. This property is typically established by
bounded input
Lyapunov-type arguments [54].
Next, we consider the system (6) with the output determined
from:
(8)
where is again assumed smooth.
A system (6) is said to be locally input-to-output stable if
function , a class- function , and
there exists a classpositive constants and such that for any initial state
with
and any input
with
, the solution
exists and the output
satisfies:
(9)
for all
.
It is said to be input-to-output stable if the local input-to-state
property holds for entire input and output spaces, and inequality
and any bounded input
(9) is satisfied for any initial state
.
The first term describes the (decreasing) effects of the initial condition, while the function in the second term bounds
the “amplification” of the input through the system. In the case
of square-integrable signals, the maximal amplification from
gain of
a given input to a given output is denoted as the
the system. This gain can be easily calculated, in general, only
for linear systems, where it equals the maximal singular value
(the supremum of the two-norm of the transfer function evalunorm) of the transfer
ated along the imaginary axis, or the
function. One of main goals of control design is then to minimize this gain, if the input represents a disturbance. There exist
a number of theorems relating Lyapunov and input-to-output
stability, and some of the main tools for establishing input-tooutput stability come from the Lyapunov approach. Note, however, that input-to-output stability describes global properties of
a system, so in its standard form, it is not suitable for study of
individual equilibrium sets. Input-to-output stability results are
Authorized licensed use limited to: Iowa State University. Downloaded on January 21, 2009 at 15:09 from IEEE Xplore. Restrictions apply.
KUNDUR et al.: DEFINITION AND CLASSIFICATION OF POWER SYSTEM STABILITY
1399
thus sometimes used in the stability analysis to establish Lyapunov stability results in a global sense [62]. For a more sophisticated use of the input-to-output stability concept, in which
the input-output properties are indexed by the operating equilibrium, see [63].
positive definite matrix solution; if such solution exists, it corresponds to a quadratic Lyapunov function that establishes stability of the system.
C.3 Stability of Linear Systems:
The direct ways to establish stability in terms of the preceding
definitions are constructive; the long experience with Lyapunov
stability offers guidelines for generating candidate Lyapunov
functions for various classes of systems, but no general systematic procedures. For the case of power systems, Lyapunov
functions are known to exist for simplified models with special features [56], [57], but again not for many realistic models.
Similarly, there are no general constructive methods to establish
input-to-output stability using (9) for nonlinear systems.
One approach of utmost importance in power engineering
practice is then to try to relate stability of a nonlinear system
to the properties of a linearized model at a certain operating
point. While such results are necessarily local, they are still of
great practical interest, especially if the operating point is judiciously selected. This is the method of choice for analytical
(as contrasted with simulation-based) software packages used
in the power industry today. The precise technical conditions required from the linearization procedure are given, for example,
in [62, p. 209-211]. The essence of the approach is that if the linearized system is uniformly asymptotically stable (in the nonautonomous case, where it is equivalent to exponential stability),
or if all eigenvalues have negative real parts (in the autonomous
case), then the original nonlinear system is also locally stable
in the suitable sense. The autonomous system case when some
eigenvalues have zero real parts, and others have negative real
parts, is covered by the center manifold theory; see [54] for an
introduction.
In this subsection, we consider a system of the form:
D.1 Complementarity of Different Approaches:
While Lyapunov and input/output approaches to defining
system stability have different flavors, they serve complementary roles in stability analysis of power systems. Intuitively
speaking, “input/output stability protects against noise disturbances, whereas Lyapunov stability protects against a single
impulse-like disturbance” [55, p. 103].
The ability to select specific equilibrium sets for analysis is
a major advantage of the Lyapunov approach; it also connects
naturally with studies of bifurcations [64] that have been of
great interest in power systems, mostly related to the topic of
voltage collapse. Note, however, that standard definitions like
and
(2) are not directly applicable, as both the starting set
are difficult to characterize with the
the technically viable set
norm-type bounds used in (2). An attempt to use such bounds
would produce results that are too conservative for practical use.
For outage of a single element (e.g., transmission line), and asand an autonomous
suming a known post-equilibrium set
system model, the starting set is a point; for a finite list of different outages of this type, the starting set will be a collection of
be “covered” by a
distinct points. The requirement that such
norm-type bound is not very suitable, as it would likely include
many other disturbances to which the system may not be stable,
and which are not of interest to the power system analyst. Note
also that partial stability may be very suitable for some system
models [60]. In a very straightforward example, we are typically
not interested in some states, like generator angles, but only in
their differences. The concept of partial stability is hence of fundamental importance in voltage and angle stability studies. In
such studies, we focus on a subset of variables describing the
power system, and we assume that the disregarded variables
will not influence the outcome of the analysis in a significant
way. In practice, we tend to use simpler reduced models where
the ignored variables do not appear, but conceptually we effectively use partial stability. The other key difficulty in analysis
stems from the fact that the construction of Lyapunov functions
for detailed power system models, particularly accounting for
load models, is still an open question, as we commented earlier. Because of these two reasons, the stability of power systems to large disturbances is typically explored in simulations.
Advances in this direction come from improved computer technology and from efficient power system models and algorithms;
for a recent review of key issues in power system simulations
that are related to stability analysis see [65]. In the case of power
system models for which there exist energy functions, it is posusing the so-called BCU
sible to approximate the viable set
method and related ideas; a detailed exposition with geometric
and topological emphasis is presented in [65].
The input/output framework is a natural choice for analysis of
some persistent disturbances acting on power systems (e.g., load
variations). Note, however, that conditions like (9) are difficult
(10)
which is the linearization of (1) around the equilibrium at the
origin. General stability conditions for the nonautonomous case
:
are given in terms of the state transition matrix
(11)
While such conditions [62, p. 193-196] are of little computational value, as it is impossible to derive an analytical expression
in the general case, they are of significant concepfor
tual value.
): The
In the case of autonomous systems (i.e.,
origin of (10) is (globally) asymptotically (exponentially) stable
if and only if all eigenvalues of have negative real parts. The
origin is stable if and only if all eigenvalues of have nonposhaving
itive real parts, and in addition, every eigenvalue of
a zero real part is a simple zero of the minimal characteristic
polynomial of .
In the autonomous case, an alternative to calculating eigenvalues of is to solve a linear Lyapunov Matrix Equation for a
D. Stability Definitions and Power Systems
Authorized licensed use limited to: Iowa State University. Downloaded on January 21, 2009 at 15:09 from IEEE Xplore. Restrictions apply.
1400
IEEE TRANSACTIONS ON POWER SYSTEMS, VOL. 19, NO. 2, MAY 2004
to establish in a non-local (large signal) setting, and simulations
are again the main option available today.
The two approaches of stability coalesce in the case of linear
system models. The use of such models is typically justified
with the assumed small size of the signals involved. A range of
powerful analysis tools (like participation factors [8]) has been
developed or adapted to power system models. For noise-type
disturbances, an interesting nonstochastic approach to the
worst-case analysis is offered by the set-based description
of noise detailed in [66]. Small signal analyses are a part of
standard practice of power system operation today.
D.2 An Illustration of a Typical Analysis Scenario:
In terms of the notation introduced here, a scenario leading
toward a blackout is as follows: Following a disturbance,
turns out to be unstable, and the system trajectory passes
. After actions of relays and line tripping, the
through
system splits into (mutually disconnected) components. The
,
,
post-fault equilibria in each component are
and some of them again turn out to be unstable, as their
are crossed by corresponding (sub)system
boundaries
trajectories. Note that up to this point
, stability analyses
have been performed. Then the stability assessment process
repeats on the third step and so on. In this framework under
a “power system,” we understand a set of elements that is
supplying a given set of loads, and if it becomes disconnected
(in graph-theoretic sense) at any point, we have to consider
as many newly created power systems as there are connected
components.
There exists a point of difference between theorists and practitioners that we want to comment upon here: Stability theorists tend to see a new system after the initial event (e.g., a line
switching), while practitioners tend to keep referring back to
the original (pre-disturbance) system. This is because stability
limits are specified in terms of pre-disturbance system conditions. While this is typically not a major obstacle, it points out
toward the need for a more comprehensive treatment of stability
theory for power systems as discussed in this section.
VI. SUMMARY
This report has addressed the issue of stability definition and
classification in power systems from a fundamental viewpoint
and has examined the practical ramifications of stability
phenomena in significant detail. A precise definition of power
system stability that is inclusive of all forms is provided.
A salient feature of the report is a systematic classification
of power system stability, and the identification of different
categories of stability behavior. Linkages between power
system reliability, security, and stability are also established
and discussed. The report also includes a rigorous treatment
of definitions and concepts of stability from mathematics
and control theory. This material is provided as background
information and to establish theoretical connections.
REFERENCES
[1] C. P. Steinmetz, “Power control and stability of electric generating stations,” AIEE Trans., vol. XXXIX, Part II, pp. 1215–1287, July 1920.
[2] AIEE Subcommittee on Interconnections and Stability Factors, “First
report of power system stability,” AIEE Trans., pp. 51–80, 1926.
[3] G. S. Vassell, “Northeast blackout of 1965,” IEEE Power Engineering
Review, pp. 4–8, Jan. 1991.
[4] S. B. Crary, I. Herlitz, and B. Favez, CIGRE SC32 Report: “System stability and voltage, power and frequency control,” CIGRE, Appendix 1,
Rep. 347, 1948.
[5] CIGRE Report: “Definitions of general terms relating to the stability of
interconnected synchronous machine,” CIGRE, paper no. 334a, 1966.
[6] C. Barbier, L. Carpentier, and F. Saccomanno, CIGRE SC32 Report:
“Tentative classification and terminologies relating to stability problems
of power systems,” ELECTRA, no. 56, 1978.
[7] IEEE TF Report, “Proposed terms and definitions for power system stability,” IEEE Trans. Power Apparatus and Systems, vol. PAS-101, pp.
1894–1897, July 1982.
[8] P. Kundur, Power System Stability and Control. New York: McGrawHill, 1994.
[9] V. I. Vorotnikov, Partial Stability and Control. Cambridge, MA:
Birkhauser, 1998.
[10] V. V. Rumyantsev and A. S. Osiraner, Stability and Stabilization of Motion With Respect to a Part of the Variables. Moscow, Nauka, Russia,
1987.
[11] N. Rouche, P. Habets, and M. Laloy, Stability Theory by Liapunov’s
Direct Method. New York: Springer, 1977.
[12] CIGRE Task Force 38.01.07 on Power System Oscillations, “Analysis
and control of power system oscillations,” CIGRE Technical Brochure,
no. 111, Dec. 1996.
[13] IEEE PES Working Group on System Oscillations, “Power System Oscillations,” IEEE Special Publication 95-TP-101, 1995.
[14] T. Van Cutsem and C. Vournas, Voltage Stability of Electric Power Systems. Norwell, MA: Kluwer, 1998.
[15] C. W. Taylor, Power System Voltage Stability. New York: McGrawHill, 1994.
[16] IEEE Special Publication 90TH0358-2-PWR, Voltage Stability of Power
Systems: Concepts, Analytical Tools, and Industry Experience, 1990.
[17] T. Van Cutsem, “Voltage instability: Phenomenon, countermeasures and
analysis methods,” Proc. IEEE, vol. 88, pp. 208–227, 2000.
[18] D. J. Hill, “Nonlinear dynamic load models with recovery for voltage
stability studies,” IEEE Trans. Power Systems, vol. 8, pp. 166–176, Feb.
1993.
[19] T. Van Cutsem and R. Mailhot, “Validation of a fast voltage stability
analysis method on the Hydro-Quebec System,” IEEE Trans. Power Systems, vol. 12, pp. 282–292, Feb. 1997.
[20] J. D. Ainsworth, A. Gavrilovic, and H. L. Thanawala, “Static and synchrounous compensators for HVDC transmission convertors connected
to weak AC systems,” 28th Session CIGRE, 1980, Paper 31–01.
[21] CIGRE Working Group 14.05 Report, Guide for Planning DC Links Terminating at AC Systems Locations Having Low Short-Circuit Capacities
Part I: AC/DC Interaction Phenomena, CIGRE Guide No. 95, 1992.
[22] CIGRE Working Group 14.05 Report, Interaction between HVDC convertors and nearby synchronous machines, CIGRE Brochure 119, Oct.
1997.
[23] G. K. Morison, B. Gao, and P. Kundur, “Voltage stability analysis using
static and dynamic approaches,” IEEE Trans. Power Systems, vol. 8, pp.
1159–1171, Aug. 1993.
[24] B. Gao, G. K. Morison, and P. Kundur, “Toward the development of
a systematic approach for voltage stability assessment of large-scale
power systems,” IEEE Trans. Power Systems, vol. 11, pp. 1314–1324,
Aug. 1996.
[25] D. J. Hill, P. A. Lof, and G. Anderson, “Analysis of long-term voltage
stability,” Proc. 10th Power Systems Computation Conf., pp. 1252–1259,
Aug. 1990.
[26] V. Ajjarapu and C. Christy, “The continuation power flow: A tool for
steady state voltage stability analysis,” Proc. IEEE Power Industry Computation Application Conf., pp. 304–311, May 1991.
[27] P. A. Lof, T. Smed, G. Andersson, and D. J. Hill, “Fast calculation of a
voltage stability index,” IEEE Trans. Power Systems, vol. 7, pp. 54–64,
Feb. 1992.
[28] CIGRE Task Force 38.02.14 Rep., Analysis and Modeling Needs of
Power Systems Under Major Frequency Disturbances, Jan. 1999.
[29] P. Kundur, D. C. Lee, J. P. Bayne, and P. L. Dandeno, “Impact of turbine
generator controls on unit performance under system disturbance conditions,” IEEE Trans. Power Apparatus and Systems, vol. PAS-104, pp.
1262–1267, June 1985.
[30] Q. B. Chow, P. Kundur, P. N. Acchione, and B. Lautsch, “Improving
nuclear generating station response for electrical grid islanding,” IEEE
Trans. Energy Conversion, vol. EC-4, pp. 406–413, Sept. 1989.
Authorized licensed use limited to: Iowa State University. Downloaded on January 21, 2009 at 15:09 from IEEE Xplore. Restrictions apply.
KUNDUR et al.: DEFINITION AND CLASSIFICATION OF POWER SYSTEM STABILITY
[31] P. Kundur, “A survey of utility experiences with power plant response
during partial load rejections and system disturbances,” IEEE Trans.
Power Apparatus and Systems, vol. PAS-100, pp. 2471–2475, May
1981.
[32] N. Hatziargyriou, E. Karapidakis, and D. Hatzifotis, “Frequency stability of power system in large islands with high wind power penetration,” Bulk Power Syst. Dynamics Control Symp.—IV Restructuring, vol.
PAS-102, Aug. 24–28, 1998, Santorini, Greece.
[33] IEEE Committee Report, “Guidelines for enhancing power plant response to partial load rejections,” IEEE Trans. Power Apparatus and
Systems, vol. PAS-102, pp. 1501–1504, June 1983.
[34] L. H. Fink and K. Carlsen, “Operating under stress and strain,” IEEE
Spectrum, vol. 15, pp. 48–53, Mar. 1978.
[35] R. Billinton, Power System Reliability Evaluation. New York: Gordon
and Breach, 1970.
[36] NERC Planning Standards (1997, Sept.). [Online]. Available:
http://www.nerc.com/~filez/pss-psg.html
[37] IEEE Working Group Report, “Reliability indices for use in bulk power
system supply adequacy evaluation,” IEEE Trans. Power Apparatus and
Systems, vol. PAS-97, pp. 1097–1103, July/Aug. 1978.
[38] CIGRE TF 38.03.12 Report, “Power system security assessment: A position paper,” ELECTRA, no. 175, Dec. 1997.
[39] T. E. Dy Liacco, “Control of Power Systems Via the Multi-Level Concept,” Case Western Reserve Univ., Syst. Res. Center, Rep. SRC-68-19,
June 1968.
[40] J. C. Willems, “Path integrals and stability,” Mathematical Control
Theory, J. Baillieul and J. C. Willems, Eds. New York: Springer,
1999.
[41] D. J. Hill and I. M. Y. Mareels, “Stability theory for differential/algebraic
systems with applications to power systems,” IEEE Trans. Circuits and
Systems, vol. 37, pp. 1416–1423, Nov. 1990.
[42] V. Venkatasubramanian, H. Schattler, and J. Zaborszky, “On the dynamics of differential-algebraic systems such as the balanced large electric power system,” in Systems and Control Theory for Power Systems, J.
H. Chow, P. V. Kokotovic, and R. J. Thomas, Eds. New York: Springer,
1995.
[43] F. Takens, “Constrained equations: A Study of implicit differential
equations and their discontinuous solutions,” in Structural Stability,
the Theory of Catastrophes and Applications in the Sciences, P. Hilton,
Ed. New York: Springer, 1976, pp. 143–234.
[44] W. R. Vitacco and A. N. Michel, “Qualitative analysis of interconnected
dynamical systems with algebraic loops: Well-posedness and stability,”
IEEE Trans. Circuits and Systems, vol. CAS-24, pp. 625–637, Nov.
1977.
[45] K. L. Praprost and K. A. Loparo, “A stability theory for constrained
dynamical systems with applications to electric power systems,” IEEE
Trans. on Automatic Control, vol. 41, pp. 1605–1617, Nov. 1996.
[46] S. R. Sanders, J. M. Noworolski, X. Z. Liu, and G. C. Verghese, “Generalized averaging method for power conversion circuits,” IEEE Trans.
Power Electronics, vol. 6, pp. 251–259, Apr. 1991.
1401
[47] A. M. Stankovic and T. Aydin, “Analysis of unbalanced power system
faults using dynamic phasors,” IEEE Trans. Power Systems, vol. 15, pp.
1062–1068, July 2000.
[48] J. S. Thorp, C. E. Seyler, and A. G. Phadke, “Electromechanical wave
propagation in large electric power systems,” IEEE Trans. Circuits and
Systems–I: Fundamental Theory and Applications, vol. 45, pp. 614–622,
June 1998.
[49] R. A. DeCarlo, M. S. Branicky, S. Pettersson, and B. Lennartson, “Perspectives and results on the stability and stabilizability of hybrid systems,” Proc. IEEE, vol. 88, no. 7, pp. 1069–1082, July 2000.
[50] T. S. Lee and S. Ghosh, “The concept of stability in asynchronous distributed decision-making systems,” IEEE Trans. Systems, Man, and Cybernetics–B: Cybernetics, vol. 30, pp. 549–561, Aug. 2000.
[51] H. Nijmeijer and A. J. van der Schaft, Nonlinear Dynamical Control
Systems. New York: Springer, 1990.
[52] V. D. Blondel and J. N. Tsitsiklis, “A survey of computational
complexity results in systems and control,” Automatica, vol. 36, pp.
1249–1274, 2000.
[53] A. N. Michel, K. Wang, and B. Hu, Qualitative Theory of Dynamical
Systems, 2nd ed. New York: Marcel Dekker, 2001.
[54] H. K. Khalil, Nonlinear Systems, 2nd ed. Englewood Cliffs, NJ: Prentice-Hall, 1996.
[55] J. C. Willems, The Analysis of Feedback Systems. Cambridge, MA:
MIT Press, 1971.
[56] M. A. Pai, Energy Function Analysis for Power System Stability. Norwell, MA: Kluwer, 1989.
[57] A. A. Fouad and V. Vittal, Power System Transient Stability Analysis
Using the Transient Energy Function Method. Englewood Cliffs, NJ:
Prentice-Hall, 1991.
[58] R. K. Miller and A. N. Michel, Ordinary Differential Equations. New
York: Academic, 1982.
[59] W. Hahn, Stability of Motion. New York: Springer, 1967.
[60] J. L. Willems, “A partial stability approach to the problem of transient
power system stability,” Int. Journal Control, vol. 19, no. 1, pp. 1–14,
1974.
[61] A. N. Michel, Qualitative Analysis of Large Scale Dynamical Systems. New York: Academic, 1977.
[62] M. Vidyasagar, Nonlinear Systems Analysis, 2nd ed. Englewood
Cliffs, NJ: Prentice-Hall, 1993.
[63] D. J. Hill and P. J. Moylan, “Connections between finite gain and
asymptotic stability,” IEEE Trans. on Automatic Control, vol. AC-25,
pp. 931–936, Oct. 1980.
[64] J. Guckenheimer and P. Holmes, Nonlinear Oscillations, Dynamical
Systems, and Bifurcations of Vector Fields. New York: Springer, 1983.
[65] H.-D Chiang, “Power system stability,” in Wiley Encyclopedia of
Electrical Engineering, J. G. Wester, Ed. New York: Wiley, 1999, pp.
104–137.
[66] F. Paganini, “A set-based approach for white noise modeling,” IEEE
Trans. on Automatic Control, vol. 41, pp. 1453–1465, Oct. 1996.
Authorized licensed use limited to: Iowa State University. Downloaded on January 21, 2009 at 15:09 from IEEE Xplore. Restrictions apply.